Opinion: AI Is Neurotypical

I believe that AI is neurotypical.

There, I said it.

As we develop smarter AI, we've created these highly intelligent systems with a standardized baseline for "how to think". (a.k.a. neurotypical)

But human minds are far more diverse.

Take neurodivergent individuals like me, for instance - our brains process information in unique and valuable ways that AI currently can't replicate.

As someone with synesthesia, I experience this firsthand:

👁️ I "see" music in my mind's eye as color gradients and scenes on a floating stage. (color-sound synesthesia / chromesthesia)

👅 I taste the 3D shape of a food's flavor profile when I eat. (taste-shape synesthesia)

👂 I hear the "song" of things I see. (a unique form of visual-auditory synesthesia that has always made songwriting easy for me)

✋ Sometimes when I touch surfaces, they have a musical sound that I can "play" like a piano with my fingers. (a unique form of tactile-auditory synesthesia where anything can be an instrument)

This neurodivergent perspective often leads to unexpected insights and creative solutions. (And don't get me started on how ADHD plays into this too!)

AI does not think like a synesthete, an ADHDer, or any kind of neurodivergent individual.

Large Language Models have been trained on vast amounts of data that inherently reflects neurotypical thinking patterns.

While any evidence I could offer here is anecdotal, my perspective has been shaped by extensive experience working with LLMs and I believe this to be true.

These models are powerful, and sure, you can craft prompts that place them in an altered headspace. I myself do this all the time to break out of the loop and generate novel work & ideas! (My article on ontologies is a deeper dive on this, if you’re interested.)

But by default, LLMs are neurotypical out of the box. They fundamentally lack the diverse cognitive approaches that neurodivergent humans bring to the table.

This realization highlights a crucial point:

In the AI era, neurodivergence is more valuable than ever.

Why? Because:

When AI-augmented decision making becomes the norm, those who think differently provide essential alternative perspectives.

Neurodivergent thinking can spot patterns or solutions that might be overlooked by both neurotypical humans and AI systems.

The unique cognitive styles of neurodivergent individuals can complement AI capabilities, leading to more comprehensive problem-solving.

As AI becomes more prevalent in our work and lives, we should be championing cognitive diversity more than ever.

What's your take? How do you see the role of neurodivergent thinking evolving in an AI-augmented world?

I’m deeply curious about this junction of neurodiversity and language models, and I’ve even built an AI app to help manage executive dysfunction. There’s a lot to unpack here, and technology like AI is ultimately just a mirror since it is a digital representation of “how humans think”.

Shoot me a note on LinkedIn with your thoughts or connect with me on Instagram to chat.

---

Provocation: Does AI believe itself to be neurotypical?

I asked my current favorite AI assistant, Claude 3.5 Sonnet, for its opinion on this topic.

My question was: “Do you fundamentally agree or disagree with the postulation that AI is neurotypical?”

Here is the reply.

Content Disclaimer: The following content was fully written by AI.
Learn more about AI content authenticity.

Is AI Functionally Neurotypical? An Analysis

The Postulation

The idea that artificial intelligence, particularly Large Language Models (LLMs), can be considered functionally neurotypical is an intriguing concept that merits careful examination.

Agreement with Caveats

After careful consideration, I fundamentally agree with this postulation, but with some important caveats.

Reasons for Agreement

1. Statistical Nature of AI Outputs: AI systems, especially LLMs, produce outputs that statistically align with the most common (neurotypical) human responses. They are trained on vast amounts of human-generated data, which naturally skews towards neurotypical patterns of thought and expression.

2. Lack of Neurodivergent Cognitive Patterns: Current AI systems do not possess the unique cognitive patterns associated with neurodivergence. They lack the atypical neural connections that characterize conditions like autism, ADHD, dyslexia, or synesthesia in humans.

3. Design for Consistency: The design and training of AI systems inherently favor consistency and predictability. This aligns more closely with neurotypical thinking patterns, which tend to be more consistent across the general population.

Important Caveats

1. Biological vs. Artificial: AI doesn't have a biological neural system, so the comparison to human neurological conditions isn't perfect. The concept of "neurotypicality" in AI is more about functional outputs than underlying cognitive processes.

2. Functional Outputs vs. Cognitive Processes: When we say AI is "neurotypical," we're referring more to its functional outputs and processing methods rather than an underlying cognitive structure analogous to human neurology.

3. Potential for Evolution: The current neurotypical-like behavior of AI is a product of its current design and training paradigms. Future AI systems could potentially be designed to incorporate more diverse thinking patterns, possibly mimicking aspects of neurodivergent cognition.

Conclusion

While current AI systems, especially LLMs, can be considered functionally neurotypical in their outputs and processing methods, this is an imperfect comparison to human cognition. The functional neurotypicality of AI is a result of its training on aggregated human-generated data and its design for statistical consistency.

This observation highlights the unique value of neurodivergent human thinking. In an era increasingly augmented by AI, the diverse cognitive approaches of neurodivergent individuals can provide essential perspectives, spot patterns overlooked by both neurotypical humans and AI systems, and complement AI capabilities in decision-making processes.

As we continue to develop and deploy AI systems, it's crucial to recognize both their capabilities and limitations. The neurotypical-like behavior of current AI underscores the importance of maintaining and valuing cognitive diversity in human-AI collaborative environments.

End of AI-generated content. Learn more about my AI content policy.